All times are UTC+01:00




Post new topic  This topic is locked, you cannot edit posts or make further replies.  [ 155 posts ]  Go to page Previous 1 2 3 4 5 616 Next
Author Message
PostPosted: Tue Aug 21, 2018 11:30 pm 
Site Admin
User avatar

Joined: Mon Mar 17, 2008 1:40 am
Posts: 11186
Stereo Tool 9.03 BETA006


Pfew, this was a really really difficult one.

I've been working on an update of the Stokkemask filter. For most people here, this won't be very useful (but: see below). In some places though (the Netherlands and Belgium) this is a mandatory thing that stations have to comply to. For most processors, the only solution is to either broadcast in mono or reduce modulation to about 68 kHz. Stereo Tool, Breakaway, and the Omnia 7/9/9sg are - to my knowledge - the only processors that can enforce Stokkemask compliance.

So what's new? Well, the existing filter was kinda sloppy, but worse, had more effect on the audio than what was necessary. Mainly on stereo separation, but it could also make the sound a bit "grainy". The grainy effect is gone now, and the effect on stereo separation is reduced by about 40% in extreme cases (so it should be more in non-extreme cases). And the filter is much stricter than before.

Why would this be interesting for people outside of the Netherlands and Belgium? Well... the whole idea of this mask is that it improves reception. And now, you can turn it on with much less effect on the audio than before (the effect was mainly noticeable when the audio was very loud and stereo was wide).

WARNINGS:
IF YOU HAVE TO BE STOKKEMASK COMPLIANT, PLEASE DO NOT USE THIS VERSION ON-AIR BEFORE READING THE FOLLOWING:
1. The new filter uses FAR MORE cpu than before. Because of that, by default, existing presets will start up in "LEGACY" mode. Which is however kinda broken right now (I will fix it tomorrow to make it equal to the old version). So, right now, if you've carefully calibrated and measured your signal, please avoid this version for now.
2. The new filter works best if you boost the clipper precision to the maximum value (18).
3. We still have to add a final filter that detects and removes any remaining filter overshoots.
4. ALWAYS be aware of this: Different measurement devices will show slightly different Stokkemask graphs. Either measure with the same device that your regulatory agency uses, or stay slightly below 75 kHz modulation.

For now, to test the new mode, go to Stokkemask and disable Legacy mode, then set the clipper strictness to (or close to) 18.



Windows 32 bit:
Windows stand alone: https://www.stereotool.com/download/ste ... 03-006.exe
Winamp DSP: https://www.stereotool.com/download/dsp ... 03-006.exe
VST: http://www.stereotool.com/download/vst_ ... 03-006.dll

Windows 64 bit:
Windows stand alone: https://www.stereotool.com/download/ste ... 03-006.exe
VST: https://www.stereotool.com/download/vst ... 03-006.dll

Latest MicroMPX:
Windows MicroMPX web based decoder: https://www.stereotool.com/download/Mic ... 03-006.exe
Windows MicroMPX web based encoder: https://www.stereotool.com/download/Mic ... 03-006.exe
Windows MicroMPX command line enc/dec: https://www.stereotool.com/download/Mic ... 03-006.exe
Windows libsndfile-1.dll (needed for MicroMPX decoder and command line version, place in same directory): MicroMPX command line enc/dec: https://www.stereotool.com/download/libsndfile-1.dll
ARM MicroMPX is coming in a few hours
ARM32 MicroMPX web based decoder: https://www.stereotool.com/download/Mic ... ETA903-006
ARM32 MicroMPX web based encoder: https://www.stereotool.com/download/Mic ... ETA903-006
ARM32 MicroMPX command line enc/dec: https://www.stereotool.com/download/Mic ... ETA903-006
ARM64 MicroMPX web based decoder: https://www.stereotool.com/download/Mic ... ETA903-006
ARM64 MicroMPX web based encoder: https://www.stereotool.com/download/Mic ... ETA903-006
ARM64 MicroMPX command line enc/dec: https://www.stereotool.com/download/Mic ... ETA903-006



CHANGES:
  • Stok: Remove overlap shortcut for stricter result (increases CPU).
  • Stok: Measure mask in every stage for stricter result (increases CPU).
  • Stok: Apply measured results in every stage for stricter result (increases CPU).
  • Stok: Replace L-R reduction in final stages by total volume reduction.
  • Stok: Figure out how to adjust shape at different sample rates -> Found correct Gaussian window sigma value instead, applied for both processing and display (which were set differently before!)
  • Stok: Replace "Looseness (sounds better)" by something that moves the mask up and down.
  • Stok: 1e stage only inner
  • Stok: More steps at end, needs to be both tighter and have less effect (is that last thing still needed???)
  • Stok: Try reducing the audio level in the final stage - lowering highs has proven useless at that point. --> SEEMS TO WORK, but only do this as the last stage!
  • MicroMPX SWEEP


    OLD CHANGES:
  • Fixed a crash in the "Limit instead" code (Advanced Clipper).
  • Removed (moved to Extreme Tweaker for possible future testing) some Stokkemask settings that had a bad effect on compliance. Good settings are 50% / 30% / 40 dB (without the 40 dB it should be fne too, but there's a bigger chance at overshoots at the bottom of the mask. 100% for the first value increases overshoots!!!!! 39 and 38 should be safe as well - but I haven't had time to measure those separately anymore. Anyway in the older versions this value was set to 0....... Setting it higher may cause an audio quality reduction.
  • Crash fixed
  • Stokkemask: Try I/Q Clipper mode, highs are coming out slightly too loud, due to LPF? Might be good enough... -> Idea doesn't work
  • Stokkemask: We don't care about CPU load, try adding more stages -> No effect
  • Stokkemask: Default back to 50% for safety - that way we don't have to worry about breaking older versions. And we still get 90% of the improvement that way.
  • More experimental stokkemask settings added.
  • Stokkemask: Improved further (looks about equally strict, but I can squeeze over 1 dB more anti-phase content through than with 9.02). NOTE: SETTINGS ARE DEFAULTING TO 9.02 BEHAVIOR BECAUSE I DON'T KNOW YET IF THE SIGNAL IS EQUALLY STRICT, AND I DON'T WANT SOMEONE TO GET FINED FOR HAVING STOKKEMASK OVERSHOOTS. To get the new behavior, set "New looseness 1" higher (50% should definitely be safe, I'm not sure about 100%), "New looseness 2" to 30%, and "Side level" to 40.
  • Improved Stokkemask - output is stricter (closer to the mask), so it has less impact on stereo separation. (*)
  • Memory usage reduced by about 12 MB

    (*) We need a measurement to make sure it's as tight as before. It probably is - but some testing is needed to make sure. I can measure it with my own analyser and with MpxTool, but neither gives a perfectly accurate result. Most importanty, I would like to know if setting the new "New looseness (experimental)" slider to 100% (0% is equal to previous versions) has any effect on compliance. Below 100% there's not really any difference in what I'm measuring, at 100% there is a small difference. ---> RESULT: 100% is bad, keep at 50. Then it's ok.


    PLANNED:
  • API: Audio signal VU meter, MPX output level setting, info about whether stream is good NOW.
  • SST/uMPX interface
  • Stok: Add post filter
  • Improve clipper sound 1-6 kHz area
  • DrawWave draws outside of screen in FM analyzer
  • The BIMP
  • Sneak peek
  • Video
  • Add Kantar support BUSY
  • Stok: Move start point to later stage?
  • Update Paul Thompson and Bojcha presets
  • Include MicroMPX - for real now [22]
  • Add MicroMPX plugin licensing [22]
  • MicroMPX: highs idea BUSY -> Find beta testers TEST STOKKEMASK EFFECT TOMORROW
    -
  • RDS urgent issue
  • Increase 60%
  • Cntrct IN
  • Cntrct TM
  • Stokkemask: Measure TOMORROW
    -
  • Show warning when using MME with wrong sample rate
  • Test preset that causes buffer to run empty - why? --> Settings issue, maybe add a popup for this situation (buffer is slowly filling up or running empty)...
  • Bug report: Linux version uses 100% CPU on one thread, even in bypass mode (must be sound card thread?) -> Not here. Not sure what's going on.....
  • Test 64 bit sound card exception handling in ProppFrexx - that wasn't the problem... apparently
    -
  • Declipper: Put startup in system tray back in Declipper (at least in the Winamp plugin version).
  • RDS: EON: Support for other type of AF frequencies?
  • RDS: Support for AF method B
  • RDS: RT end at \n
  • Brian, NAB
  • MOBO
  • uMPX encoder/plugin: Sound card speed message
  • uMPX decoder: Add multicast subscribe
  • Add MagicRDS AF method B support


Top
   
 
PostPosted: Wed Aug 22, 2018 2:15 am 

Joined: Sat Sep 05, 2015 6:12 pm
Posts: 21
1000 NaN errors on the ProcessingChain, glitches in the audio. Beta 006 32bit can be considered as failed build, tested on Windows 10 x32 version, Windows 7 x32 version, and Windows 8.1 x32 version.
I'm in bed now, so can't take screenshots at the moment :)

9.02 BETA018 was all fine and stable.


Top
   
 
PostPosted: Wed Aug 22, 2018 10:01 am 
Site Admin
User avatar

Joined: Mon Mar 17, 2008 1:40 am
Posts: 11186
:(

Confirmed. It's caused by Stokkemask "legacy mode" - as soon as I turn that off it's ok again. In debug mode it's ok as well - it only fails in the release build :( . Will look into it today.


Top
   
 
PostPosted: Wed Aug 22, 2018 12:11 pm 
User avatar

Joined: Tue Mar 17, 2009 2:56 pm
Posts: 4151
It uses enormous amount of CPU with strictness 18.
Actually even with strictness 8. Also more cpu usage when graph is OFF.

_________________
control point
control point2


Top
   
 
PostPosted: Wed Aug 22, 2018 12:53 pm 
User avatar

Joined: Tue Mar 04, 2014 10:17 am
Posts: 65
Is it posible to use GPU for processing instead of CPU


Top
   
 
PostPosted: Wed Aug 22, 2018 3:10 pm 
Site Admin
User avatar

Joined: Mon Mar 17, 2008 1:40 am
Posts: 11186
Quote:
It uses enormous amount of CPU with strictness 18.
Actually even with strictness 8. Also more cpu usage when graph is OFF.
I know. It sounds better though and is stricter. And I still need to try to make it work as well with lower strictness settings. So, if you have the CPU power available you can run it. (I can ALMOST run it at 150% at the normal 1 thread + multicore mode on my pc).

I'm going to revert legacy mode to what it was before, that should fix both the NAN's issue and the performance. And make sure that anyone who has it setup correctly doesn't have to worry about it.


Top
   
 
PostPosted: Wed Aug 22, 2018 4:39 pm 
User avatar

Joined: Tue Mar 17, 2009 2:56 pm
Posts: 4151
On my PC i can run it. But it's i7 4790K at 4.7GHz and still with 50%. Old E8500 at 3.6GHz ccant work with this at all. And with legacy stoke makes 85% cpu usage already.

_________________
control point
control point2


Top
   
 
PostPosted: Wed Aug 22, 2018 4:51 pm 
Site Admin
User avatar

Joined: Mon Mar 17, 2008 1:40 am
Posts: 11186
Quote:
On my PC i can run it. But it's i7 4790K at 4.7GHz and still with 50%. Old E8500 at 3.6GHz ccant work with this at all. And with legacy stoke makes 85% cpu usage already.
Plan:

Step 1: Offer both legacy and new mode (legacy with same behavior and CPU load as before).

Step 2: Optimize new mode to work with lower CPU load, maybe slightly less strict.

In the end I hope that even if your CPU can't handle the extra CPU load required for the more precise filtering, the new method will still give a better sound than the old. But for now, if your CPU can't handle it, stick with the old version...


Top
   
 
PostPosted: Wed Aug 22, 2018 10:15 pm 

Joined: Mon Jun 08, 2015 7:28 pm
Posts: 94
Location: Belgium
Hi,
I have try your last version but I receive: Exception Occured in ProcessingChain Thread (Exception 0xc0000005 at address 0x7ff65202c3ad, read/write=0, r/w location=0x263c26eadc0).
Offset-info: 0x7ff6528ab409

Restart will be attempted.

and i restart and loop on this error.

Only composite clipper and FM transmitter is enabled in processiong (ST on transmitter side) ;-)


Top
   
 
PostPosted: Wed Aug 22, 2018 10:41 pm 
User avatar

Joined: Tue Mar 04, 2014 10:17 am
Posts: 65
Quote:
Hi,
I have try your last version but I receive: Exception Occured in ProcessingChain Thread (Exception 0xc0000005 at address 0x7ff65202c3ad, read/write=0, r/w location=0x263c26eadc0).
Offset-info: 0x7ff6528ab409

Restart will be attempted.

and i restart and loop on this error.

Only composite clipper and FM transmitter is enabled in processiong (ST on transmitter side) ;-)
I also recieve a lot of theese warnings (even when I do not use Legacy mode in Stockmark) but it works anyway they are just anoying

also is it possible to use GPU for procesing instead of CPU


Top
   
 
Display posts from previous:  Sort by  
Post new topic  This topic is locked, you cannot edit posts or make further replies.  [ 155 posts ]  Go to page Previous 1 2 3 4 5 616 Next

All times are UTC+01:00


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Limited